Smart Home Device Brand Perception on Twitter

Introduction

Smart home devices serve as intelligent personal assistant and could be used as a home automation system.

The global smart home market is growing at a fast pace, with an expected value of USD 53.45 billion in 2022. Amazon Alexa, Google Home, and Apple HomePod are the most famous smart home devices.

Alexa: a 9-inch tall cylinder speaker controlled by a cloud-base voice assistant released by Amazon in Nov 2014.
Google Home: a brand of smart speakers developed by Google in late 2016, which enable users to speak voice commands to interact with services through Google’s intelligent personal assistant.
Apple Home: early 2018, Apple will also be joining the fray with the Siri-powered HomePod.

Get twitter data

Set up twitter

–api_key <- “RrPTCIryCswqP35AFeanh73cV”

–api_secret <- “w4JEhIMHsEArgaandqAtTqhc610kRvenP8e8fdXfSRCnuLVsn9”

–access_token <- “927638841925742593-3NjhdkxNIJRbGQ8PPDIMIpFskIMFpPd”

–access_token_secret <- “Qfd1mKmPbi0E6O6cmblVG4Fmi6CQceG1CtbyQKR28nAAY”

–setup_twitter_oauth(api_key, api_secret, access_token, access_token_secret)

Get tweets and transform into data framework

–Alexatweet <- searchTwitter(“Amazon Alexa”, n=3000, lang=“en”, since=“2014-01-01”)

–Alexatweetdf <- twListToDF(Alexatweet)

–GoogleHometweet <- searchTwitter(“Google Home”, n=3000, lang=“en”, since=“2014-01-01”)

–GoogleHometweetdf <- twListToDF(GoogleHometweet)

–AppleHometweet <- searchTwitter(“Apple HomePod”, n=3000, lang=“en”, since=“2014-01-01”)

–AppleHometweetdf <- twListToDF(AppleHometweet)

Store tweets data

–write.csv(Alexatweetdf,“Alexa.csv”)

–write.csv(GoogleHometweetdf,“GoogleHome.csv”)

–write.csv(AppleHometweetdf,“AppleHome.csv”)

Text Analysis

Import the stored data and unnest tokens the tweets texts

Alexadf<-read.csv("Alexa.csv")
GHdf<-read.csv("GoogleHome.csv")
AHdf<-read.csv("AppleHome.csv")

Alexatext<-select(Alexadf,text)
Alexatext$text<-as.character(Alexatext$text)
Alexatext<- Alexatext %>% 
  unnest_tokens(word, text)

GHtext<-select(GHdf,text)
GHtext$text<-as.character(GHtext$text)
GHtext<- GHtext %>% 
  unnest_tokens(word, text)

AHtext<-select(AHdf,text)
AHtext$text<-as.character(AHtext$text)
AHtext<- AHtext %>% 
  unnest_tokens(word, text)

Remove stop words

mystopwords<-c("https","rt","htt","votedougjones")

mystopwords<-as.data.frame(mystopwords)
colnames(mystopwords)<-"word"
mystopwords$word<-as.character(mystopwords$word)

Alexa<-Alexatext %>%
  anti_join(stop_words) %>% 
  anti_join(mystopwords) %>%
  count(word, sort = TRUE) %>%
  filter(str_detect(word,"^[a-z']+$"))

## Joining, by = "word"
## Joining, by = "word"

GH<-GHtext %>%
  anti_join(stop_words) %>% 
  anti_join(mystopwords) %>%
  count(word, sort = TRUE) %>%
  filter(str_detect(word,"^[a-z']+$"))

## Joining, by = "word"
## Joining, by = "word"

AH<-AHtext %>%
  anti_join(stop_words) %>% 
  anti_join(mystopwords) %>%
  count(word, sort = TRUE) %>%
  filter(str_detect(word,"^[a-z']+$"))

## Joining, by = "word"
## Joining, by = "word"

Amazon Alexa EDA

As we can see, the tops words of Alexa are Alexa, Amazon, music, wake, echo, win, chance, amp, follow,dorischristmas, alarms, and gift.

These words reflect the main features of Alexa perceived by customers:

it’s a smart echo produced by Amazon;
people use it to play music and also use as the alarm to wake them up;
people consider it to be a good Christmas gift and there are some marketing campaigns where people can win Alexa as a free gift.

Besides, we also found the words like, Alex, Jones, and CIA. As Alexa Jones grilled an Amazon Echo on his show, asking Alexa — the device’s personal assistant — about its connections to the CIA.

Google Home EDA

The top words for Google Home are:home, google, max, smart, speaker, mini, follow, sign, volume and incredible.

These words reflect how people perceive Google Home:

Google Home is a mini smart speaker;
there is a new product called Google Home Max with bigger volume;
there are ongoing marketing events on Twitter, where you follow a company and you can get Google Home;

Besides, incredible implies that people have a positive impression of Google Home.

Apple Home EDA

As we can see, the top words of Apple Home are apple, homepod, win, giving, enter, gleamapp, winner, lucky home, cnet, siri, audio.

From these top words we can see how customer perceive Apple HomePod:

Gleam is doing Apple HomePod Giveaway campaign, where people can win Apple HomePod as a free gift;
cnet published a news that Apple HomePod is delayed until early 2018;
Apple HomePod is powered by Siri and allows multi-speaker audio streaming;

Besides, we can see people are comparing Apple HomePod with Google Home.

People are also discussing the price of Apple HomePod, as it has the higher price compared with Alexa and Google Home.

Alexa Sentiment analysis

## Joining, by = "word"

## We recommend that you use the dev version of ggplot2 with `ggplotly()`
## Install it with: `devtools::install_github('hadley/ggplot2')`

We did sentiment analysis to see how people’s comment on Alexa,overall, there are more positive words about Alexa.

we can see there are some negative words with high frequency, such as hard and lying, which may indicate people have some difficulties of properly using Alexa;

While win is a positive word with very high frequency related to Alexa, we can see people are interested in the marketing campaign of Alexa.

Google Home Sentiment analysis

## Joining, by = "word"

## We recommend that you use the dev version of ggplot2 with `ggplotly()`
## Install it with: `devtools::install_github('hadley/ggplot2')`

Overall, there are more positive words about Google Home. We can see there are some negative words, such as damn and loud, indicating people feel google home is loud.

While smart and incredible are positive words with high frequency, we can see more people think Google Home is smart and incredible

Apple Home Sentiment analysis

## Joining, by = "word"

## We recommend that you use the dev version of ggplot2 with `ggplotly()`
## Install it with: `devtools::install_github('hadley/ggplot2')`

Unlike Amazon Alexa and Google Home, Apple HomePod has more positive words, such as lucky, win and winner, which is related to the marketing campaign.

Still, there are some negative words but with lower frequency like bugs, delay and wrong.

Correlation analysis

allproduct<- bind_rows(mutate(GH, product = "GoogleHome"),
                       mutate(AH, product = "AppleHome"), 
                       mutate(Alexa, product = "Alexa")) 


allpro <- allproduct %>% 
  mutate(proportion = round(n / sum(n),5)) %>% 
  select(-n) %>% 
  spread(product, proportion) %>% 
  gather(product, proportion, "GoogleHome":"AppleHome") %>%
  na.omit(proportion)


cor.test(data = allpro[allpro$product == "GoogleHome",],
         ~ proportion + Alexa)

## 
##  Pearson's product-moment correlation
## 
## data:  proportion and Alexa
## t = 2.1505, df = 852, p-value = 0.0318
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
##  0.006420462 0.139870952
## sample estimates:
##       cor 
## 0.0734746

cor.test(data = allpro[allpro$product == "AppleHome",], 
         ~ proportion + Alexa)

## 
##  Pearson's product-moment correlation
## 
## data:  proportion and Alexa
## t = 1.696, df = 396, p-value = 0.09067
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
##  -0.0134912  0.1817008
## sample estimates:
##        cor 
## 0.08491946

allpro2 <- allproduct %>% 
  mutate(proportion = round(n / sum(n),5)) %>% 
  select(-n) %>% 
  spread(product, proportion) %>% 
  gather(product, proportion, "AppleHome":"Alexa") %>%
  na.omit(proportion)

cor.test(data = allpro2[allpro2$product == "AppleHome",], 
         ~ proportion + GoogleHome)

## 
##  Pearson's product-moment correlation
## 
## data:  proportion and GoogleHome
## t = 2.3107, df = 564, p-value = 0.02121
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
##  0.01454096 0.17783612
## sample estimates:
##        cor 
## 0.09684018

As they are similar products, we did correlation analysis.

Our null hypothesis test: there is no correlation between the sentiment of two products.

The significance level is 0.05.

We found there is small positive correlation between the sentiment of Alexa and apple home.

No significant correlation between Alexa and Google Home nor between Google Home and Apple Home.

Conclusion

We can see there are both similarity and difference in people’s perception to Alexa, Google Home, and Apple Home.

The features of alarm and music player of Alexa are perceived by the customer, while there are a lot of other features that Amazon can put efforts to promote among customers, such as controlling smart home devices, read news and audiobooks, and make calls.There are some negative aspects of Alexa, like a connection with CIA. Amazon could clarify on public and build a good image of Alexa.

While Google Home has a better reputation in the market, people recognize it as a smart and incredible home device. But people are also complaining the loud volume of Google Home. Google could improve the volume control and solve some technical bugs of google home. There are not many discussions about the specific features of Google Home. Google could make some efforts to promote the specific features of Google Home.

Apple Home hasn’t been launched yet, but people are discussing it a lot, such as the delay of its release, the price and the techniques including Siri and audio streaming. And People are comparing it with Google Home. So it’s import for Apple to promote some special features that separate Apple Home from Google Home and Alexa.

All these three products have online marketing campaigns which attract customers. Christmas would be a good opportunity to promote these products.